The KITTI Vision Benchmark Suite

In this work, we propose RoI-Aware Grids Pooling Networks for 3D
object detection, i.e., RagNet3D. Specifically, we devise RoI-
View Prediction Networks, simplified as RVPNet, to predict grid-
wise RoI-View residuals. Then, we also propose Couple-View
Fusion Networks, denoted as CoVFNet, to fuse the view-dependent
features and view-independent features, learned from predicted
RoI-View and 3D convolution networks respectively.

TBD

@article{chen4979473ragnet3d,
title={Ragnet3d: Learning Distinguishable Representation for Pooled Grids in 3d Object Detection},
author={Chen, Jiaxin and Han, Yuehui and Yan, Zhiqiang and Qian, Jianjun and Li, Jun and Yang, Jian},
journal={Available at SSRN 4979473}
}